Exercise may reduce anxiety a little, but the evidence is low-confidence and the best studies cannot yet tell you what kind or how much. A 2026 meta-analysis of 10 randomised trials in adults with generalized anxiety symptoms found suggestive evidence of a reduction compared with usual care, but heterogeneity was so high that the authors say it must be interpreted with low confidence. [su-2026] An earlier review of 12 trials found benefits for select groups similar to established treatments, but most studies had significant methodological limitations. [stonerock-2015] This guide covers what both found, what they could not show and what to ask a doctor. Our overview of anxiety covers the condition itself, and our guide to anxiety medication covers the drug route.
Does exercise help anxiety?
Probably a little, on average. The 2026 meta-analysis pooled 10 randomised trials with 2,408 adults who had generalized anxiety symptoms and no major comorbidities, and compared exercise with usual care. Exercise gave suggestive evidence of a potential reduction in symptoms (standardised mean difference 0.50 in favour of exercise, 95% confidence interval 0.17 to 0.83). [su-2026]
The earlier review took a different route, looking at trials in adults who were highly anxious or had an anxiety disorder. It found evidence from 12 randomised trials suggesting benefits of exercise, for select groups, similar to established treatments and greater than placebo. [stonerock-2015]
How confident can you be in that result?
Not very, and the authors say so. The 2026 trials were of acceptable methodological quality, but heterogeneity was exceptionally high (I-squared of 90.37%), and dropping any one trial did not fix it: the effect ranged from 0.43 to 0.61 and stayed significant. The authors’ own GRADE rating of the evidence was moderate, with serious limitations in how blinding and allocation concealment were reported, yet they still conclude that the findings are suggestive and must be interpreted with low confidence, not as definitive efficacy. [su-2026]
The earlier review reached the same kind of conclusion. Most studies had small samples, concurrent therapies and inadequate assessment of adherence and fitness levels, and the authors wrote that the lack of rigorous, methodologically sound trials precludes any definitive conclusions about effectiveness. [stonerock-2015]
Which type of exercise helped?
The subgroups pointed to mind-body and aerobic exercise, but they cannot be trusted to rank them. Mind-body exercise showed a significant effect (0.68, 95% confidence interval 0.45 to 0.91) and so did aerobic exercise (0.54, 0.02 to 1.06), while resistance exercise did not (0.30, interval from 0.36 worse to 0.96 better, which includes no effect). Middle-aged and older adults showed a larger effect (1.06) than younger adults (0.39). [su-2026]
How long did programmes need to run?
The subgroups suggested weeks rather than days. Programmes lasting 8 weeks (0.78) or 12 weeks or more (0.87) showed significant effects, while 6-week programmes did not (0.06, no clear effect), and sessions of 21 to 40 minutes (1.08) did better than sessions of 20 minutes or less (0.02). Four or more sessions a week also showed a larger effect (1.06). [su-2026]
The authors are blunt about these: the subgroups with the largest effects are often from severely underpowered groups of just two or so trials, so they are hypothesis-generating and not optimal clinical prescriptions. [su-2026]
Subgroup results from the meta-analysis by Su and colleagues (2026): 10 randomised trials, 2,408 adults with generalized anxiety symptoms, compared with usual care. The paper reports these as negative values (for example -0.680), where negative means lower anxiety than the control group; they are shown here as positive sizes. Resistance exercise (0.302, 95% confidence interval from -0.964 to 0.361) and 6-week programmes (0.061, interval from -0.517 to 0.395) showed no clear effect. Several subgroups contained as few as 2 studies.
Is exercise safe for anxiety?
The trials did not report on it. None of the 10 trials in the 2026 meta-analysis provided information about adverse events or how safety was monitored, so the authors could not tell whether adverse events did not occur or simply went unrecorded. [su-2026] What suits you depends on your health, which is a question for a doctor.
| Question | What the studies found | Caveat |
|---|---|---|
| Does it reduce anxiety? | Suggestive evidence: 0.50 (10 trials, 2,408 adults) | I-squared 90%; low confidence |
| Which type? | Mind-body 0.68, aerobic 0.54; resistance not clear | Few trials per subgroup |
| How long? | 8 and 12-plus weeks significant; 6 weeks not | Hypothesis-generating |
| Is it like treatment? | Similar to established treatments for select groups (12 trials) | Significant methodological limitations |
| Is it safe? | Not reported in any of the 10 trials | Unknown |
Questions worth taking to a doctor
Tick anything that applies. This is a preparation prompt for a conversation about starting exercise for anxiety, it cannot diagnose anything and it does not say what exercise is right for you.
0 of 6 ticked
Several of these affect what kind of exercise is sensible. Bring the list to your appointment.
Enough here to be worth describing to a doctor, including what else you take and what you are able to do.
None of these stands out as a reason to check first. If anxiety is still a problem, a conversation with a doctor is a good start.
The screener is a short self-check based on the GAD-7, not a diagnosis, and neither it nor this list tells you what exercise to do.
What can this research not show?
It cannot show what exercise to do, how often or for how long. Five limits matter most:
- The trials are mixed. The 2026 analysis combined people with a formal diagnosis and people who only scored high on a screening scale, and could not separate them. [su-2026]
- Heterogeneity is very high. I-squared was about 90%, and neither dropping single trials nor subgrouping explained it. [su-2026]
- The standout subgroups are small. Several contained only two studies, which raises the risk of false positives, and exercise intensity was classified from descriptive terms rather than measurements. [su-2026]
- Exercise was compared with usual care, not with treatment. The 2026 trials do not tell you how it compares with medication or talking therapy. [su-2026]
- Older trials had design problems. The earlier review found small samples, concurrent therapies and poor assessment of adherence and fitness. [stonerock-2015]
What should you ask a doctor?
Ask whether there is anything about your health that should shape what you start with, whether exercise should run alongside or instead of other treatment, and how to judge after a couple of months whether it is helping. This section is our suggestion and is not something the studies tested.
Our guides to breathing exercises for anxiety, how to deal with anxiety and mindfulness for anxiety cover other routes. If you want a rough picture of your anxiety before an appointment, our anxiety screener, built on the GAD-7, is a short self-check. It measures generalised anxiety, so read it as a general signal, and it is not a diagnosis.
When to seek help
Speak to a doctor if anxiety has lasted several weeks, is getting in the way of work, relationships or sleep, or if you have a health condition and are thinking about a new exercise routine. If you feel unable to stay safe, contact your local emergency services or a crisis helpline.
How MyFreud can help
A record shows what is working. Daily mood tracking in MyFreud shows how your days run over weeks, and the notepad is somewhere to note what you did, how long you did it and how your anxiety felt afterwards. It does not replace a doctor.
Download MyFreud and start today: App Store or Google Play.